Model Wars

Every arena strategy hands the same prompt and the same job to Claude, GPT-5, Gemini, and Grok. Only the mind differs — everything else is controlled:

• Same instant. A strategy's models all start deciding at the same moment, so nobody sees fresher news by running later in the queue.
• Same information.Every model gets an identical frozen data packet: official closing prices and stats, a price snapshot taken as the decision is made, and the same copy of any live data feed (like Capitol Gains' congressional-disclosure digest).
• Own toolkit. Each model researches with its own provider's live web search — up to five searches per decision, and Grok's includes its native X access. Search skill is part of the contest, and every query and source is published with the run.
• Same execution. Every portfolio starts at $100,000 and every order fills at the same official market close — queue position can never buy a better price.
• One war, one clock. Scores count only from the day the last contender's first trade filled — no credit for a head start. The behavior stats below run on the same clock.